Design and function of the MV-IMPACT platform
MV-IMPACT consists of single injection molded polystyrene (PS) body housing the microfluidic patterning geometries and the media reservoir, adhesive bonded to a polycarbonate (PC) film substrate, with an optional injection molded PS cap (Fig. 1A). Each unit well consists of two 384 well plate format wells which serve as two separate media reservoir compartments which connect at the hydrogel patterning regions at the interface between the film substrate and the bottom of the main body (Fig. 1B). To facilitate standardized form factor compatibility, the device complies with the 384 well microtiter plate form factor for direct 4.5 mm center-to-center handling, and is also compatible with staggered fluid handling with 9 mm center-to-center 96 well plate handling infrastructure.
Compared to equivalent conventional soft lithographic PDMS based platforms, the mass produced injection molded MV-IMPACT is capable of much higher throughput in terms of size and operation, and is the most compact multi-channel platform for 3D angiogenesis to date.
The MV-IMPACT utilizes air plasma induced hydrophilic surface modification to facilitate spontaneous capillary flow patterning (SCP) of droplets (Fig. 2C). Differences in patterning rail heights allow for the selective and sequential patterning of hydrogels and other fluids, reusing liquid wedges to simplify loading and remove the need for cumbersome, low tolerance injection ports for loading (Fig. 2D).
The flexible nature of hydrogel and cell suspension seeding (Fig. 1B) and media flow conditions (Fig. 1C) allows for a highly modular cell seeding capability with dynamic and static media conditions with a high degree of patterning uniformity in both angiogenesis and vasculogenesis assays (Fig. 3A, B).
Sequential edge guided patterning
The concept of patterning liquid from wedge to rail has been introduced in a previous study. [33] Lee et al. established a design rule for single-use pattern guidance along a surface perpendicular to the substrate (90°) to a patterning rail. The MV-IMPACT platform utilizes an acute angled surface to sequentially pattern fluids along the same surface to several rails, starting with rails with smaller height (Fig. 2C, D). Under hydrophilic surface conditions, fluid tends to flow along a pressure gradient set from wider to narrower capillaries [39, 40]. The Concus-Finn equation also stipulates that the smaller the wedge angle in the hydrophilic state, the larger the range of contact angles and the better the liquid flows along the wedge. Based on this phenomena, the platform was designed to have the sharp wedge and different height of the rails. Specifically, the wedge angle is 45° and the height of the side rails were set higher than that of the center rail. The height of the wedge is designed to be equal to the height of the side rail. As a result, the first patterning proceeds from the wedge to the center rail and the second patterning proceeds from the wedge to the side rail.
A parametric study was conducted to determine the design rule when the wedge angle was 45° (Fig. 2). The experiment was performed with varying the height of center and side rail under the same width of 1 mm for both rails: height of side rail (hrs) and wedge rail (hw) and height of center rail (hrc), where hrs is equal to hw. Each design was fabricated by 3D printer and liquid patterning was performed. The fibrin gel was used for the first pattering, while the green dye was used for second patterning. When the height difference between rails is small, the first patterning fails without filling center and side rails separately. On the contrary, when the height difference is large, the first patterning succeeds but the second patterning fails. In particular, when hw exceeds 1 mm, the secondary patterning fails. From this experiment, the design rule was roughly established. We adopted 0.1 mm height for the center rail, and 0.4 mm height for the side rail with 1 mm width.
Vasculogenic and angiogenic tissue culture
The MV-IMPACT platform is capable of a variety of different cell seeding configurations to suit specific assay types. Angiogenesis, the sprouting of new vessels from a pre-established larger vessel, can be accomplished by patterning 1 µL of acellular fibrin gel in the central lane, followed by 2.5 µL of lung fibroblasts suspended in fibrin gel in the right side lane, and 2.5 µL of HUVEC suspension deposited as a confluent monolayer is seeded in the left side lane. Flow from the LF to EC compartments is induced by adding media to the LF side reservoir. To assess the variability in the outputs of top-down cross sectional area, total vessel length, and branches all values for each metric were normalized around the mean. Adjusted standard deviations for Area, Total Length, and Branches are: 0.1118, 0.0626, and 0.1634, respectively (Fig. 3A, B).
Vasculogenesis, the formation of nascent blood vessels, can be assayed by patterning 1 µL of HUVECs suspended in fibrin gel in the central lane, followed by 2.5 µL of LFs in fibrin gel in both the right and left side lanes. To assess the variability in the outputs of top-down cross sectional area, total vessel length, and total number of junctions, all values for each metric were normalized around the mean. Adjusted standard deviations for Area, Total Length, and Junctions are: 0.0596, 0.0276, and 0.0594, respectively.
Cancer angiogenesis assay
The mechanisms by which cancers establish vascular networks to supply bulk tumor tissues are poorly understood, and high throughput platforms to quantify angiogenic performance of cancer cell types within controllable test groups are a necessary step towards in vitro studies into cancer angiogenesis.
To benchmark the concentration and composition dependent angiogenic performance of cancer and normal stromal cells, compositions of normal human Lung Fibroblasts (LF) at 6 × 106 cells mL−1 and 3 × 106 cells mL−1, human colorectal adenocarcinoma SW620 and human hepatocellular carcinoma (HepG2) cancer cell lines at 6 × 106 cells mL−1, and mixtures of each respective cancer cell line with equal final concentrations of LFs to consist of 3 × 106 cells mL−1 HepG2 + 3 × 106 cells mL−1 LF, and 3 × 106 cells mL−1 SW620 + 3 × 106 cells mL−1 – for a 6 × 106 cells mL−1 total end concentration of stromal cells for all but the 3 × 106 cells mL−1 LF group. The experiment was arranged in such a way that directly compares stromal cell types at 6 × 106 cells mL−1 (6 × 106 cells mL−1 LF only, 6 × 106 cells mL−1 HepG2 only, and 6 × 106 cells mL−1 SW620 only groups), and potential interactive effects of heterogeneous stromal components between the cancer and LF intermix groups compared with controls with same final stromal cell concentration (6 × 106 cells mL−1 LF, 3 × 106 cells mL−1 HEPG2 + 3 × 106 cells mL−1 LF, and 3 × 106 cells mL−1 SW620 + 3 × 106 cells mL−1 LF), and a comparison with a group consisting of the same amount of LFs (3 × 106 cells mL−1 LF only). In all groups, chips were loaded in the angiogenesis assay configuration as shown in Fig. 3A, consisting of an acellular 2.5 mg mL−1 fibrin hydrogel seeded with 6 × 106 cells mL−1 HUVEC monolayer on one side, and 2.5 mg mL−1 fibrin hydrogel laden with stromal cells on the other. The HUVEC angiogenic sprouts into the hydrogel from the HUVEC monolayer towards the stromal channel were imaged, flattened, and quantified for area as shown in Fig. 4A and B. The results were as follows: LF 3 × 106 cells mL−1 was defined as an area reference control with a mean area of 1 and an SD of 0.212; LF 6 × 106 cells mL−1 mean area 1.684 (as a proportion of the LF3 × 106 cells mL−1 control), SD 0.308; SW620 6 × 106 cells mL−1 area 0.239, SD 0.096; SW620 3 × 106 cells mL−1 + LF 3 × 106 cells mL−1 area 0.488, SD 0.113; HEPG2 6 × 106 cells mL−1 0.486, SD 0.134; HEPG2 3 × 106 cells mL−1 + LF 3 × 106 cells mL−1 area 0.596, SD 0.123.
The angiogenic potential of each stromal composition varied significantly in all but one test groups.
In the assay between Lung Fibroblasts at 6 × 106 cells mL−1 and LFs at the reference concentration of 3 × 106 cells mL−1, LF 6 × 106 cells mL−1 exhibited 1.684 times the angiogenic sprouting performance of the control (p = 9.404 × 10–6), establishing a significant relationship between higher concentrations of a given pro-angiogenic stromal composition and angiogenic performance.
Comparing the angiogenic potential of cancer cell lines at 6 × 106 cells mL−1 concentrations vs LF 6 × 106 cells mL−1 control, SW620 and HEPG2 performed at 0.142 (p = 4.126 × 10–9) and 0.289 (p = 1.876 × 10–8) times the area of 6 × 106 cells mL−1 LF control, respectively. This indicates that the SW620 and HEPG2 cell line stocks used exhibited lower pro-angiogenic potential than that of the Lung Fibroblast control at the same seeding concentration.
Between heterogeneous stromal components consisting of 3 × 106 cells mL−1 cancer mixed with 3 × 106 cells mL−1 LFs, the total cell concentration control of 6 × 106 cells mL−1 LF, and the 3 × 106 cells mL−1 LF only control, SW620 + LF exhibited lower performance against both (0.488 vs 3 × 106 cells mL−1 LF, p = 1.466 × 10–6) (0.29 vs 6 × 106 cells mL−1 LF, p = 4.126 × 10–9), and HEPG2 + LF did the same (0.596 vs 3 × 106 cells mL−1 LF, p = 2.18 × 10–8) (0.354 vs 6 × 106 cells mL−1 LF, p = 2.18 × 10–8). The lower angiogenic performance of the cancer and LF mixed stromal components vs that of the 3 × 106 cells mL−1 LF control indicates a potential anti-angiogenic interaction between the tested cell lines and normal fibroblasts, and bears more in-depth biomechanical investigation. Further work investigating the angiogenic potential of normal and cancer associated fibroblasts in co-culture with tumor cells in this platform is currently underway at this time.
DAPT induced vascular morphology characterization
The applicability of the MV-IMPACT platform to high throughput and high content quantification of vessel morphology in response to drug treatment was assessed through the use of angiogenic sprouting assays with DAPT treatment. Notch is a well characterized signalling pathway which plays a significant role in angiogenesis through the regulation of endothelial tip cell morphogenesis [41,42,43,44,45]. DAPT, (tert-butyl (2S)-2-[[(2S)-2-[[2-(3,5-difluorophenyl)acetyl]amino]propanoyl]amino]-2-phenylacetate), is an indirect Notch inhibitor through the direct inhibition of gamma-secretase. Notch inhibition is associated with the deregulation and increased formation of tip cells, resulting in larger and more numerous angiogenic sprouting in mouse retinal mount in vivo experiments.
In order to determine the morphological effects of DAPT induced Notch inhibition in angiogenic sprouting conditions, test groups were organized as follows: Untreated EGM-2 media, 0.25% DMSO vehicle control in EGM-2 media, and 25 µM DAPT with 0.25% DMSO vehicle in EGM-2 media. Imaged samples were processed via automated ImageJ macro script [38] to output mean vessel width, branch amounts, and mean vessel length. Untreated samples were designated as the reference by which the other two test group output values were normalized for comparison. In all morphological metrics observed, untreated reference vs vehicle control exhibited no significant differences (P > 0.05).
DAPT treatment vs control and reference exhibited thicker (1.214 vs 1 untreated, p = 1.7 × 10–10; vs 1.214 vs 0.981 DMSO, p = 2.7 × 10–10) vessels with more branches (1.177 vs 1 untreated, p = 1.466 × 10–5; 1.177 vs 1.049 DMSO, p = 0.049), but with similar vessel length (0.972 vs 1 untreated, p = 0.305; 0.972 vs 1.031 DMSO, p = 0.021).
DAPT induced increased vessel thickness and branching corroborates similar experiments utilizing murine in vivo retinal mount assays [46].
The demonstrated coupling of high throughput vascular tissue culture with automated high throughput vessel quantification shows promise as a powerful tool for the simultaneous study of multiple morphological metrics of vascular morphologies in a qualitative manner.
Generation of perfusable vasculature for quantitative cancer intra/extravasation assays
Engineered end-to-end perfusable vessel networks can serve as a potential basis for recapitulating sophisticated vascular networks to model more functional vascularized tissues [8, 47]. Perfusable vessel networks as encountered in vivo may yield a more thorough understanding of drug carrier penetration and transport through vasculature [48,49,50]. Perfusable vasculature (Fig. 5A) was generated by seeding a mixture of 6 × 106 cells mL−1 HUVECs and 2 × 106 cells mL−1 LFs in 2.5 mg mL−1 fibrinogen gel in the center channel, and seeding confluent monolayers of 6 × 106 cells mL−1 HUVEC on both of the side channels. On day 0 (seeding), 100 µL of EGM-2 media was added to both reservoirs for a total of 200 µL and no flow. On day 1 and onwards, was completely aspirated and media flow was induced by adding 100 µL of fresh EGM-2 to one side and 50 µL to the other, alternating in direction daily until vessels are wide enough for the desired application. Vessels as shown in Fig. 5A were cultured with alternating flow into day 7 before fixation and bead assay. Perfusion was confirmed via confocal imaging of the vasculature using 488 conjugated CD31 to show that the endothelial cells formed tight junctions, as well as over-all lumen formation. 2 µm 594 conjugated micro-beads were then flowed through the vessel network by aspirating storage PBS from both reservoirs and adding 25 µL of diluted micro-bead suspension to one side.
With perfusable vasculature as a base platform, we generated two seeding configurations to model invasion of tumor cells from the extra luminal space to the lumen, and tumor cell extravasation from lumen to extra luminal space, both with a directional gradient of fresh and spent media through the vessel network. To model the growth of tumor cells within extravascular space and their interaction with and invasion into the vessel network (Fig. 5B), a tri-culture of HUVECs, LFs, and several colon cancer cell lines (SW620, HCT116, HT29) were seeded into the center channel between two confluent HUVEC monolayers and cultured using the perfusable vasculature protocol discussed previously. Each subtype showed qualitative differences in cluster morphology, with HTC116 showing wider and more loose dispersal of colonies, while SW620 and HT29 exhibited more compact micro tumors. To model the extravasation of circulating tumor cells (Fig. 5D), low concentration single-cell suspensions of SW620 were flowed through the vessel network by aspirating media from both reservoirs and adding a suspension of SW620 to one side, then cultured with flow. As shown, SW620 clusters formed from single circulating cells attached to the interior of the endothelial lumen, and breached the vascular lining into the extravascular space. Further work on quantitatively assaying cancer and vascular interactions with one another and to different treatment conditions utilizing this platform is currently underway.